UC San Diego

Digital image compositing is a fundamental operation of computer graphics. For the sake of performance, current implementations shun many important software qualities such as brevity, readability, and extensibility. This thesis presents a novel implementation that uses a powerful programming environment to maintain these software qualities while still allowing for high performance. The approach is based on run-time code generation, using a simple object pipeline architecture to manage software complexity. The resulting implementation, named "Jitblt," is compared to an existing, popular compositing library. The results show that large gains in software quality can be had, but further work is necessary to make this new approach practical